This paper is concerned with how collaborative virtual environments can be structured in order to enable greater scalability and yet maintain a richness of communication. Based on a survey of current work in this area a number of important issues and design criteria are drawn out. In this context a recent extension to the spatial model of interaction called third party objects is described which can be used to develop a variety of spatial structures which have different effects on mutual awareness and communication. These structures include nested regions and zones, dynamic data districts and potentially mobile crowds of participants. The paper also discusses how these structures can be exploited by an underlying network architecture based on a dynamically evolving hierarchy of multicast groups and describes the MASSIVE-2 system which implements it.